IBM 5655-DB2 Manuel D’Utilisation
148
Storage Management with DB2 for OS/390
discarding "foreign overhead" from the target study. However, values related to
this foreign overhead must be preserved because interactions exist on resource
access inside the same computing perimeter. The first step is an overall analysis.
Moreover, this first overview allows locating some missing data and finding
information from other sources. In this case, some input lacking from the RMF
cache analysis reports was compensated by IXFP.
this foreign overhead must be preserved because interactions exist on resource
access inside the same computing perimeter. The first step is an overall analysis.
Moreover, this first overview allows locating some missing data and finding
information from other sources. In this case, some input lacking from the RMF
cache analysis reports was compensated by IXFP.
An efficient approach is to build a spreadsheet, doing some consolidation
between the cache subsystem activity and device activity reports. In this case
study, 46 pairs of LCU level RMF reports were reviewed before selecting 9 of
them as relevant. There was no foreign overhead to take into account. Figure 67
on page 148 shows the result of this preliminary step. RMFPP was not used
because we just needed to capture global activities at the LCU and at Storage
Group levels. The report source line is either "crr" for cache activity, "da" for
device activity source, or "crr/da". The fields to review are as follows:
between the cache subsystem activity and device activity reports. In this case
study, 46 pairs of LCU level RMF reports were reviewed before selecting 9 of
them as relevant. There was no foreign overhead to take into account. Figure 67
on page 148 shows the result of this preliminary step. RMFPP was not used
because we just needed to capture global activities at the LCU and at Storage
Group levels. The report source line is either "crr" for cache activity, "da" for
device activity source, or "crr/da". The fields to review are as follows:
• CUID/ADDR is the CU-ID field of the cache subsystem activity report header
cross-checked with the DEV NUM of the direct access device activity report.
• SSID is the SSID field of the cache subsystem activity report header.
• LCU is the LCU field of the direct access device activity report.
• rmf_rate is the LCU total device activity rate from the direct access device
activity report.
• crr_rate is the aggregate LCU I/O rate field from the cache subsystem device
overview report. These values are higher than the values RMF captured
because control units count at the command level (locate record command)
rather than at the channel program level for RMF. Moreover, there is a missing
value for the two last LCUs because of some operator interaction during the
measurement interval. The fact the "crr" and the "rmf" values are close to each
other shows there is no data sharing with other LPARs.
because control units count at the command level (locate record command)
rather than at the channel program level for RMF. Moreover, there is a missing
value for the two last LCUs because of some operator interaction during the
measurement interval. The fact the "crr" and the "rmf" values are close to each
other shows there is no data sharing with other LPARs.
In most cases, a preknowledge of the environment shortens this pruning process.
Figure 67. Reducing the RMF Data to Analyze
REDUCTION
reports crr/da
crr
da
da
crr
field
CUID/ADD
SSID
LCU
rate
rate
unit
ssch / sI/O / sec
RVA_1
1st LCU
2B00
0088
0046
15.8
18.2
2nd LCU
2B40
0089
0047
8.8
10.1
3rd LCU
2B80
008A
0048
14.1
15.9
4th LCU
2BC0
008B
0049
7.9
missing
Tot
rva1
46.6
44.200
RVA_2
1st LCU
2C00
2007
004A
3.7
4.2
2nd LCU
2C40
2008
004B
14.4
15.9
3rd LCU
2C80
2009
004C
14.5
16.8
4th LCU
2CC0
200A
004D
11.7
missing
Tot
rva2
44.4
36.900
System
71C0
603C
0055
7.0
22.1